SCaNME: Location tracking system in large-scale campus Wi-Fi environment using unlabeled mobility map
Introduction
Location tracking in wireless network has attracted significant interest in recent years. Knowledge of people’s real-time locations and related routine activities make many promising applications for industrial and public uses (Horng et al., 2011, Zhang et al., 2013) for childcare, elderly health care, emergency rescue, as well as management in warehouse and modern buildings (Cura, 2010, Dey et al., 2011). Moreover, by the year 2014, the GPS technology could yield more than 2.5 billion dollars device market and over 12.7 billion dollars are expected to be reported by location-based service (LBS) market at that time (Ashbrook & Starner, 2003).
GPS and the Cellular network are two established infrastructures for location tracking, but they cannot work well in some urban or indoor environments due to the poor quality of the received signal from satellites or base stations. Serious attenuation and multipath interference exists in challenging radio propagation environments, such as buildings, tunnels, and subway stations (Anisetti et al., 2011, Wong et al., 2009). Therefore, a large amount of effort has been made to track people’s locations by alternative technologies, including the radio frequency identification (RFID) (Silva & Goncalves, 2009), ultrasonic wave (UW) Yoshiga et al., 2012, ZigBee (Goncalo & Helena, 2009), Bluetooth (Mair & Mahmoud, 2012), ultra-wideband (UWB) Steiner & Wittneben, 2010, and infrared ray (IR) Chen et al., 2010. However, these technologies require additional costs of infrastructure and maintenance, great effort for fingerprint or landmark calibration, frequent signaling between the mobile users and infrastructure, and rigorous application scenarios.
Based on the previous work on the location tracking in Wi-Fi environments (Ouyang et al., 2012, Ouyang et al., 2010, Zhou et al., 2013, Zhou et al., 2013), we can find that the Wi-Fi technology has the advantages of high accuracy, low cost, and widespread deployment. Moreover, the Wi-Fi modules can be integrated in most of current personal mobile devices, such as the laptops, cell phones, tablets, and personal digital assistants (PDAs). Therefore, after the world’s pioneering Wi-Fi location tracking system, RADAR, was invented by Microsoft Research in 2000 (Bahl & Padmanabhan, 2000), many universities and institutes began to develop a variety of prototype and commercial systems to track people’s locations (Castro et al., 2003, Garcia-Valverde et al., 2013, Swangmuang and Krishnamurthy, 2008, Youssef and Agrawala, 2008).
So far, there are generally four types of location tracking techniques by using the Wi-Fi networks: the time or time difference of arrival (TOA or TDOA) (Golden and Bateman, 2007, Schwalowsky et al., 2010), angle of arrival (AOA) (Wong et al., 2007, Wong et al., 2008), propagation model (Ahn and Yu, 2008, Emery and Denko, 2007), and received signal strength (RSS)-based fingerprinting (Bahl and Padmanabhan, 2000, Castro et al., 2003, Swangmuang and Krishnamurthy, 2008, Youssef and Agrawala, 2008). By TOA or TDOA, strict time synchronization of access points (APs) and mobile stations within a network is required (Golden and Bateman, 2007, Schwalowsky et al., 2010). However, multipath fading and shadowing in target environment make it difficult to obtain the exact arrival time of signal and propagation distance. In other words, the results of location tracking obtained by the combination of log-distance path loss model with triangulation algorithm cannot be accurate. For AOA technique, it is suggested that the smart antenna is used at the receiver to track locations by measuring the angles of arrival signal (Wong et al., 2007, Wong et al., 2008). In all, the TOA or TDOA and AOA techniques heavily rely on the sophisticated propagation model and extra hardware to achieve location tracking.
In comparison, the Wi-Fi RSS-based fingerprinting is preferred for location tracking because there is no requirement for any extra devices besides the widely deployed Wi-Fi networks in public. In general, there are two phases involved in the tracking process, the off-line phase and on-line phase (Fang and Lin, 2010, Zhou et al., 2012). In the off-line phase, the fingerprints from the RSS measurements at reference points (RPs) and the corresponding physical coordinates are recorded to construct a radio map. Then, in the on-line phase, the people’s locations are estimated by matching the pre-stored fingerprints with the new RSS data. The major drawback of RSS-based fingerprinting is the intensive labor cost for the fingerprint calibration.
To solve this problem, we propose a novel location tracking system called SCaNME which can adaptively construct an unlabeled mobility map and track people’s locations and paths without any calibration of fingerprints. The research presented in this study is an extension of our previous work (Zhou et al., 2013). Different from the conventional radio map, the unlabeled mobility map is represented by a graph of RSS clusters associated with non-zero transition probabilities. Meanwhile, by using the Allen’s logics on the temporal relations of time stamped RSS clusters, the people’s activities can be recorded from the mobility map. In this paper, we mainly focus on the activity recording for the people from the same organization. Our proposed SCaNME is beneficial to the future user-centric ubiquitous LBS which requires not only the knowledge of people’s real-time location information, but also the understanding of people’s routine activities. In the environments where the people’s movement patterns have been recorded, we can use SCaNME for loss finding, such as Find my phone at iphone. On the other hand, for healthcare applications, by tracking the movement of elderly people, the nursing center can monitor the activities of their care-receivers, or the families of the elder can track their locations anytime on-line in cloud. In all, the activity recording-based LBS could make our life easier and securer. Another application of SCaNME system is to describe the layout of anonymous environments adequately and adaptively. The off-line RSS training involved in conventional fingerprint-based location tracking systems requires significant time cost and periodic RSS management. To guarantee the efficiency of RSS training, a rough floor plan could be essential to the creation of off-line RSS database. However, in many anonymous environments, the floor plans are probably not available to the users for the security or privacy reasons. To solve this problem, the mobility map constructed by SCaNME system can be used to describe the target environmental layout in an effective manner. The SCaNME consists of the off-line map training phase and the on-line location tracking phase. The objectives of the off-line map training phase are to construct an unlabeled mobility map and record the people’s motion behaviors. During the on-line location tracking phase, the people’s locations and paths are tracked in a real-time manner.
The remaining of this paper is organized as follows. Section 2 summarizes some related work on Wi-Fi location tracking and describes the purpose of mobility map construction in SCaNME system. Section 3 explains the steps of tracking people’s locations by SCaNME system. In Section 4, we discuss the way to record the people’s activities from the unlabeled mobility map by Allen’s logics. Section 5 provides some numerical results collected from the experiments in a campus-wide Wi-Fi network. In Section 6, we conclude the paper and present our future work.
Section snippets
Related work
In the recent decade, the proliferation of ubiquitous LBS demand has fostered the research of Wi-Fi location tracking in both industry and academia. The fundamental model for the RSS-based location tracking was built by Microsoft Research’s RADAR system (Bahl & Padmanabhan, 2000). In RADAR system, the ideas of RSS-based fingerprinting and radio propagation model for the location tracking are introduced, which also become the two leading schemes to locate the people’s positions in Wi-Fi
Architecture of SCaNME location tracking
There are two phases involved in SCaNME system, the off-line mobility map construction phase and the on-line path reconstruction phase. An overview of our proposed SCaNME system is illustrated in Fig. 1. After the mobility map is constructed, we track the people’s positions by matching the newly collected RSS data into the pre-stored RSS samples. Details of each step in SCaNME system are described below.
Activity learning by Allen’s logics
In addition to the steps of location tracking in SCaNME system, the SCaNME also instantaneously learns the person’s activities from the previously reconstructed paths, in order to accurately and effectively locate the tracking LPs on the paths. To meet this goal, we use Allen’s logics as a new metric to investigate the person’s motion activities conveyed from the mobility map. Therefore, the probability of a LP being the tracking LP should also depend on the person’s prior knowledge of
Data collection
The efficiency and effectiveness of the location tracking by the SCaNME system is verified in one part of the HKUST campus (http://publish.ust.hk/univ/campu). The volunteers carrying the Samsung GT19100 Android phones collected 2017 Wi-Fi RSS measurements (of dimension 588) following their routine activities on Monday. They used the same wireless card vendor during both the off-line and on-line phases to reduce the ambiguity of displayed RSS values (Kaemarungsi, 2006, Mengual et al., 2010). In
Conclusion
Location tracking in large-scale environments is addressed as a passport to the all-weather and seamless LBSs in future. How to use widely-deployed Wi-Fi networks to perform the highly accurate and cost efficient location tracking remains to be an open issue. The conventional Wi-Fi RSS-based location fingerprinting involves a plethora of time and laboring cost for fingerprint recording. However, in SCaNME, we first use the sporadically recorded Wi-Fi RSS measurements to create a mobility map in
Acknowledgement
The authors wish to thank the reviewers and editors for their valuable comments and useful corrections. This work was done when Mu Zhou was a Post-doctoral Research Assistant at The Hong Kong University of Science and Technology (HKUST). This work was supported in part by the National Natural Science Foundation of China (61301126, 61304197), National Science and Technology Major Project of China (2011ZX03006003007, 2012ZX03006002003), Fundamental and Frontier Research Project of Chongqing (
References (51)
A parallel local search approach to solving the uncapacitated warehouse location problem
Computers & Industrial Engineering
(2010)- et al.
Enhancing WLAN location privacy using mobile behavior
Expert Systems with Applications
(2011) - et al.
Clustering-based location in wireless networks
Expert Systems with Applications
(2010) - et al.
An effective location fingerprint model for wireless indoor localization
Pervasive and Mobile Computing
(2008) - et al.
Theoretical entropy assessment of fingerprint-based Wi-Fi localization accuracy
Expert Systems with Applications
(2013) - Ahmad, U., Gavrilov, A., Lee, S. Y., & Lee, Y. K. (2006). Modular multilayer perceptron for WLAN based localization. In...
- Ahn, H. S., & Yu, W. (2008). Wireless localization networks for indoor service robots. In Proceedings of IEEE/ASME MESA...
- Alasti, H., Xu, K., & Dang, Z. (2009). Efficient experimental path loss exponent measurement for uniformly attenuated...
Maintaining knowledge about temporal intervals
Communications of the ACM
(1983)- et al.
Map-based location and tracking in multipath outdoor mobile networks
IEEE Trans. Wireless Communications
(2011)
Using GPS to learn significant locations and predict movement across multiple users
Personal and Ubiquitous Computing
Indoor tracking and navigation using received signal strength and compressive sensing on a mobile device
IEEE Transactions on Mobile Computing
Indoor location system based on discriminant-adaptive neural network in IEEE 802.11 environments
IEEE Transactions on Neural Networks
A dynamic system approach for radio location fingerprinting in wireless local area networks
IEEE Transactions on Communications
A fuzzy logic-based system for indoor localization using WiFi in ambient intelligent environments
IEEE Transactions on Fuzzy Systems
Sensor measurements for Wi-Fi location with emphasis on time-of-arrival ranging
IEEE Transactions on Mobile Computing
Cited by (29)
Stalking the stalkers – detecting and deterring stalking behaviours using technology: A review
2017, Computers and SecurityCitation Excerpt :However devices conducting this randomisation can often still be tracked by fingerprinting the lower-layer attributes of the device, by using active attacks or by keeping track of sequence numbers over time (Freudiger, 2015; Vanhoef et al., 2016). There exist many publications covering the tracking of moving targets using multiple fixed sensors (Li et al., 2014, 2015; O'Connor, 2013; Zhou et al., 2014) which make that research primarily useful for private companies who wish to learn more about their customers' movements, or law enforcement agencies who wish to conduct mass surveillance of an area. These entities have the financial resources and legal authority to mount permanent or semi-permanent sensors over a large area.
An integrated architecture for future studies in data processing for smart cities
2017, Microprocessors and MicrosystemsComprehensive analysis of distance and similarity measures for Wi-Fi fingerprinting indoor positioning systems
2015, Expert Systems with ApplicationsCitation Excerpt :Wi-Fi is a good choice for indoor positioning technology due to the ubiquity of wireless access points(WAPs), embedded Wi-Fi connectivity in modern mobile phones, and the use of pre-existing Wi-Fi network infrastructure(s). The IPSs based on Wi-Fi fingerprinting are preferred to those based on the propagation model, angle of arrival, time of arrival and time difference of arrival because they do not require any very specialized hardware, line-of-sight to the emitter and knowing emitter’s location to operate (Yu et al., 2014; Zhou et al., 2014a). Fig. 1 shows an example of fingerprinting where a smartphone detects six WAPs.
Characterizing activity sequences using profile Hidden Markov Models
2015, Expert Systems with ApplicationsCitation Excerpt :The characterization of each group’ s travel behavior and its similarities or differences with other groups’ conducts could provide important insights into the design of marketing strategies that are tailored to each particular group (Shoval & Isaacson, 2007b). Regarding context-awareness services and location tracking systems where knowledge of individuals’ real-time as well as routine locations is vital to provide user-centric assistance (García-Sánchez, GonzáLez, Mora, & Prieto, 2013; Horng, Chen, Ferng, Kao, & Li, 2011; Lee & Cho, 2013; Zhou et al., 2014), the individual-pHMMs characterizing each individual’s travel behavior could offer considerable values in the enhancement of the personalized location-based services. Besides the multiple potential applications of the profile method, this study also makes important contributions from a theoretical point of view.
Efficient Services Management in Libraries using AI and Wireless techniques
2014, Expert Systems with ApplicationsCitation Excerpt :Our proposal does not pretend to propose any new method on how tags should be placed but how AI can be used to coordinate and monitor all the information about users, rooms and books. Some approaches use Wi-Fi for people tracking (Zhoua et al., 2014) but instead, we have decided RFID as the indoor location system (Tesoriero, Tebar, Gallud, Lozano, & Penichet, 2010). The reason was mainly influenced by the RFID infrastructure already available in libraries for books, that could be extended also for people.
Building a validation measure for activity-based transportation models based on mobile phone data
2014, Expert Systems with ApplicationsCitation Excerpt :Besides the initial goal of building a new benchmark measure for activity-based models, the proposed method for stop location identification and subsequent actual-travel-sequence derivation can be employed in a wide range of applications where location path information of cell phone users plays a central role. The applications do not only include travel behavior and transportation modeling related research, such as mobility pattern discovery (Do & Gatica-Pereza, 2013; Schneider et al., 2013), transportation modeling and traffic analysis (Angelakis et al., 2013; Berlingerio et al., 2013; Calabrese et al., 2011), and urban planning (Becker et al., 2011; Jiang, Joseph Ferreira, & Gonzalez, 2012), but they also cover context-awareness services where user-centric assistance is provided based on users’ specific location and activity context (GarcíA-SáNchez, GonzáLez, Mora, & Prieto, 2013; Lee & Cho, 2013), and location tracking systems where knowledge of individuals’ real-time locations and related routine activities is used in tools that provide support for industry, childcare, elderly health care and emergency rescue (Horng, Chen, Ferng, Kao, & Li, 2011; Zhang, Liu, Jian, & Guan, 2013; Zhou et al., 2014). Despite the multitude of possible applications, there are also challenges that are pertinent to the data, as acknowledged by some of the existing studies (e.g. Calabrese et al., 2011; Jiang et al., 2013; Rose, 2006).